📄 Grammaticalization of the Amharic tense system

Summary

Łykowska examines how the tense, aspect and modality (TAM) system of Amharic came about through grammaticalization. She draws on attested historical texts, such as the 14th–16th c. "Songs of Kings", compares Amharic with Tigrinia and Zway, and works through glossed modern examples. Her framework is Heine's theory of complex schemas. The central claim is that the old Semitic consonant-vowel templates now do only a small part of the work. They still separate perfective from imperfective, and the plain past grew directly out of the perfective. Most other tense, aspect and modal meanings, however, are carried by auxiliary verbs and clause-level constructions that grow out of lexical verbs such as "be", "remain", "wait" and "sit". On top of Heine's Purpose and Evaluative Schemas, the author proposes a Temporal Relation (Simultaneity) Schema of the form "X acts while Y" and shows that it generates continuous, habitual, progressive and epistemic meanings. The study matters because it traces a live grammaticalization process in a Semitic language. It also argues that clause-linking structures, and not only single lexical items, can become sources of grammar.

Key points

• The modern Amharic system has four tenses: past (negger-e), present-future (yi-negr-all), remote past (negr-o nebber) and a present/present perfect (negr-o-all). All except the past are periphrastic, built with the auxiliary allé "be" or nebber.

• How much the auxiliary inflects depends on the nonfinite verb it attaches to:

  • with a prefix-inflected nonfinite (present-future), the auxiliary inflects fully;
  • with a suffix-inflected nonfinite, the auxiliary tends to freeze in a reduced 3SGM form (-all');
  • the author explains this as redundancy in agreement.

• The aspects identified are inchoative, continuous, habitual and progressive.

  • Inchoative uses auxiliaries such as jemmere "start" inside a Purpose Schema (l- + imperfect).
  • Continuity uses k'oyye/t'ebbek'e "wait", tek'emmet'e "sit" and "spend the day/night" verbs.

• The Simultaneity Schema appears in at least four structures:

  • s- "while" clauses;
  • iyye- "while" clauses;
  • a gerund + finite verb construction;
  • a verbal noun with locative be-… lay "on" plus copula.

  The last of these shows a locative marker turning temporal.

• Epistemic modality relies on hone "become" and k'erre "remain". A doubly negated k'erre construction expresses certainty or high probability. Obligation, necessity and ability are expressed with nore, the copula, asfellege "need" and chale "can", mostly through the Purpose Schema.

Interesting ideas

• I1. The paper treats a nonfinite verb plus a frozen auxiliary as one tense unit, even when material such as object pronouns sits between the two parts, rather than reading them as separate predicates.
• I2. A whole clause-linking pattern ("X while Y") can itself carry aspect, so durativity or habituality may be encoded in the construction and not in any single verb.
• I3. Two negations combine into a strong positive modal reading, as when "it will not fail to rain" means near-certainty.
• I4. Lexical verbs (wait, sit, live, remain) slide along a chain toward auxiliary status. Near-synonyms become interchangeable at that stage, but syntax can still block particular swaps.
• I5. Agreement tends to be marked only once in a compound form, so whichever part already carries the suffixal person marking lets the other part fossilize.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (tree): When converting UD dialects, chains like "has been going to" could be collapsed into a single TAM feature bundle on the head verb, so that intervening adverbs or clitics do not break tense recognition.
• I2 (world): Event ordering could flag "while"/"as"-clauses and verbs like "spend the day V-ing" or "keep V-ing" as durative overlap relations, giving the timeline explicit simultaneity intervals.
• I3 (frames): The implicature module could add rules mapping English litotes ("can't not", "not unlikely", "never fails to") to high-probability or certainty values instead of naively cancelling the negations.
• I4 (world): The compiled verb-class layer could tag English aspectual or light uses of verbs ("sit waiting", "go on", "end up") with a grammaticalization stage, so that frames assign them an aspect role rather than a full agentive event.
• I5 (tokens): The morphology tagger could include a redundancy check that expects agreement on exactly one element of an auxiliary chain and flags parses that assign conflicting person or number features.